Designing an AI Roleplay System for Contextualized Career Guidance in Mexico

Abstract

Upper secondary education students in metropolitan areas of Mexico face a persistent gap between their initial career expectations and the real conditions that influence university persistence, including commuting time, institutional costs, and family financial constraints. Traditional career guidance tools, typically based on static psychometric questionnaires, do not incorporate these contextual logistical and socioeconomic factors. This paper presents the design process of a ChatGPT-based roleplay system that simulates professional career pathways for students in the Mexican context of 2026. Rather than reporting the long-term effectiveness of the tool on career decision-making, this paper documents how the system was designed, including the iterative prompt engineering process across three major versions, a four-phase conversational architecture that builds an eight-dimension student profile and guides users through a five-stage life simulation, and a small-scale formative evaluation conducted with upper secondary students to determine whether the system could be used as intended. The contribution of this paper is not a claim of vocational effectiveness, but a reproducible methodological account of how a large language model can be configured through prompt engineering to incorporate contextual, socioeconomic, and culturally situated constraints into a career guidance experience. The formative evaluation provided a descriptive overview of changes observed among the participating students and, more importantly from a design perspective, identified a specific usability issue in the data collection procedure. This issue is discussed in relation to the existing literature on technology adoption in career guidance services.